Research Scientist, Multi-Modal Understanding & Synthesis

Meta•Redmond, WA

2d•$184,000 - $257,000

About The Position

Meta is seeking a Research Scientist to drive foundational research in multi-modal understanding, synthesis, and world models. In this role, you will advance the state of the art in building AI systems that perceive, reason across, and generate content spanning vision, language, audio, and other modalities. You will develop world models that learn rich internal representations of human behavior, enabling prediction, planning, and simulation. Collaborating with world-class researchers and engineers, you will define research directions, publish influential work, and translate breakthroughs into technologies that power Meta's next-generation AI products.

Requirements

Bachelor's degree in Computer Science, Computer Engineering, relevant technical field, or equivalent practical experience
PhD in Machine Learning, Computer Vision, Natural Language Processing, or a closely related field
6+ years of experience conducting AI research in multi-modal learning, generative models, or world models, including experience leading major research initiatives from conception through publication or production deployment
Experience implementing and evaluating multi-modal systems using deep learning frameworks such as PyTorch or TensorFlow, with proficiency in Python
Experience publishing original research in peer-reviewed machine learning or AI venues
Experience driving cross-functional technical decisions and communicating research findings and trade-offs to both research and engineering audiences through written documents and presentations
Experience with techniques spanning multiple modalities such as vision-language models, multi-modal transformers, or cross-modal representation learning

Nice To Haves

Experience developing large-scale multi-modal foundation models or vision-language models
Experience with world models, predictive learning, or model-based reinforcement learning for planning and reasoning
First-author publications at top-tier venues such as NeurIPS, ICLR, or CVPR demonstrating contributions to multi-modal learning, generative models, or world models

Responsibilities

Lead original research in multi-modal learning, developing architectures and algorithms that unify understanding and generation across vision, language, audio, and other modalities
Design and build world models that learn predictive representations of human behavior, supporting capabilities such as simulation, planning, and reasoning
Drive end-to-end research projects from problem formulation and dataset curation through model development, evaluation, and integration into real-time prototypes
Develop novel approaches for multi-modal synthesis, enabling coherent generation of images, video, text, and audio from unified representations
Establish rigorous evaluation frameworks, benchmarks, and metrics to measure progress in multi-modal reasoning and world modeling
Mentor other engineers and researchers on the team, providing technical guidance on multi-modal architectures, generative models, and research best practices
Publish research findings at top-tier peer-reviewed venues such as NeurIPS, ICLR, and CVPR to advance the broader scientific community